Adapted Multimodal BERT with Layer-wise Fusion for Sentiment Analysis

TL;DR

This paper introduces Adapted Multimodal BERT (AMB), a parameter-efficient architecture that fuses audio-visual data with text for sentiment analysis, outperforming state-of-the-art models while maintaining robustness and efficiency.

Contribution

The paper presents a novel BERT-based multimodal model using adapter modules and layer-wise fusion, enabling efficient training and improved performance in sentiment analysis.

Findings

AMB outperforms current state-of-the-art models on CMU-MOSEI.

The approach achieves a 3.4% reduction in error.

It improves 7-class classification accuracy by 2.1%.

Abstract

Multimodal learning pipelines have benefited from the success of pretrained language models. However, this comes at the cost of increased model parameters. In this work, we propose Adapted Multimodal BERT (AMB), a BERT-based architecture for multimodal tasks that uses a combination of adapter modules and intermediate fusion layers. The adapter adjusts the pretrained language model for the task at hand, while the fusion layers perform task-specific, layer-wise fusion of audio-visual information with textual BERT representations. During the adaptation process the pre-trained language model parameters remain frozen, allowing for fast, parameter-efficient training. In our ablations we see that this approach leads to efficient models, that can outperform their fine-tuned counterparts and are robust to input noise. Our experiments on sentiment analysis with CMU-MOSEI show that AMB outperforms the current state-of-the-art across metrics, with 3.4% relative reduction in the resulting error and 2.1% relative improvement in 7-class classification accuracy.